Genomic Structure and Identification of Novel Mutations in Usherin, the Gene Responsible for Usher Syndrome Type IIa

Genomic Structure

To identify and isolate bacterial artificial chromosomes (BACs) containing the USH2A gene, BAC libraries from Genome Systems and Research Genetics were screened with PCR by use of primer pairs from the critical region of the USH2A gene; these were listed in the database maintained by the Whitehead Institute for Biomedical Research/MIT Center for Genome Research. We obtained additional sequence-tagged sites by sequencing the ends of the BACs obtained from the initial BAC library screening.

PCR Conditions

Oligonucleotide primers were designed from the consensus sequence of the USH2A cDNA sequence and used for PCR and cycle-sequencing reactions. BAC clones were grown in 100 ml of Luria-Bertani broth supplemented with 12.5 mg/ml chloramphenicol overnight and purified with Qiagen tip 100 columns. Human genomic DNA was isolated from whole blood with a Puregene kit (Gentra Systems).

Primer combinations for PCR reactions with Taq Gold (Perkin Elmer) or Elongase (Gibco-BRL) were tested with human genomic DNA and BAC clones as templates. PCR reactions were as follows: 10 ng BAC or 200 ng human genomic DNA, 1× reaction buffer, 0.2 mM each dNTP, 2.0 mM MgCl₂, 0.2 mM each primer; 95°C for 10 min followed by 37 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 2 min. Elongase enzyme reactions were as follows: 10 ng BAC or 200 ng human genomic DNA, 1× reaction buffer, 0.2 mM each dNTP, 1.5 mM or 2.0 mM MgCl₂, 0.2 mM each primer for genomic templates, and 0.4 mM each primer for BAC templates; 94°C for 30 s followed by 35 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 10 min.

DNA Sequencing

PCR products were purified and concentrated prior to sequencing by Microcon 100 filters (Ambion). ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction kits (Perkin Elmer) were used to generate sequence from both purified PCR products and BAC DNA templates. PCR products were sequenced with 4 pmol Usherin-specific primers and 4 ml ready-reaction mix in a total volume of 10 ml for 25 cycles of 95°C for 10 s, 50°C for 5 s, and 60°C for 4 min. BAC DNA (500 ng–2 mg) was sequenced in 20-ml reaction volumes by 30 pmol of SP6, T7 pBELOBAC11 vector primers or USH2A primers, 8 ml ready-reaction mix in a total volume of 20 ml for 95°C for 5 min followed by 99 cycles of 95°C for 20 s, 55°C for 30 s, and 60°C for 4 min. Sequencing reactions were precipitated and resuspended in a 5:1 ratio of formamide/dye loading buffer (6 ml for PCR products and 2 ml for BAC DNA). Two microliters were loaded onto 4.8% (36-cM plates) or 5.25% (48-cM plates) PAGE-PLUS™ (Amresco) sequencing gels and run on the ABI Prism model 377 DNA sequencer according to ABI protocols. Sample files were generated from gel files by ABI Sequencing Analysis 3.0. Fractura 2.0 (ABI) was used to remove vector and ambiguous sequences. Contig assembly was performed by the AUTOASSEMBLER program (ABI).

Mutation Screening

Table 1 lists the intronic primers used in this study and the amplification sizes of their PCR products surveyed for mutations. A set of 95 anonymous control DNA samples was also screened to assess the frequency of USH2A mutations in a population unaffected by USHIIA.

Heteroduplex Analysis

PCR products amplified from samples taken from USHIIa probands and cloned control DNA were mixed, heated at 95°C for 3 min, and cooled to 25°C over a 45-min period. The reannealed reaction products were then electrophoresed through 35 cm × 43 cm × 1 mm 1× MDE™ (FMC) at 850 V for 18–24 h, stained with 1 mg/ml ethidium bromide, and visualized under ultraviolet light. Mixing with control PCR DNA prior to heteroduplex formation was done to allow detection of both homozygous and heterozygous mutations.

Direct Link Sequencing

Five-microliter PCR products were treated with 10 U of shrimp alkaline phosphatase (SAP; Amersham) for 30 min at 37°C, and the enzyme was inactivated at 80°C for 15 min. SAP was used to dephosphorylate unincorporated dNTPs prior to cycle sequencing. Sequencing reactions contained 2.8 ml of the BigDye Terminator Cycle Sequencing Ready Reaction mix (Perkin Elmer), 2.5 ml SAP-treated PCR product, and 20 pmol primer. Cycling conditions were 30 cycles of 95°C for 10 s, 55°C for 5 s, and 60°C for 4 min. Sequencing reactions were precipitated in 60% ethyl alcohol (final), dried, and resuspended in 4 ml of a 5:1 ratio of formamide/loading dye. One microliter was loaded onto 4.8% PAGE-PLUS™ (Amresco) 36-cm sequencing gels and run on the ABI Prism model 377 DNA sequencer according to ABI protocols. Sample files were generated from gel files by ABI Sequencing Analysis 3.3. The Fractura 2.2 program was used to remove ambiguous sequences and identify heterozygous base positions. Sequence data were aligned and analyzed for mutations by Sequence Navigator 1.0.1.

Comparative Analysis

A total of 32 proteins bearing laminin domain VI motifs were used to evaluate the functional significance of the three missense mutations found within domain VI of Usherin (see fig. 4 below). The first 300 amino acids from each of these protein sequences were compared with amino acids 250–550 of the Usherin protein by a progressive pairwise alignment strategy by the MULTIALIGN program (Pôle Bio-Informatique Lyonnais; Corpet 1988).

Genomic Structure

Figure 1 presents an ∼325-kb contig assembly across the region that contains the human USH2A gene. Amplification products produced by Elongase by primers specific for both USH2A cDNA and BAC DNA corresponded to seven fragments containing the following exons with their intervening introns: 1–3, 5–8, 10–11, 12–13, 17–18, 18–19, and 19–20. The intron sizes were estimated by agarose gel electrophoresis or were sequenced (IVS1, -6, -7, and -17). Attempts to amplify the following introns were unsuccessful: 3, 4, 8, 9, 11, 13, 14, 15, 16, and 20. Intron/exon boundaries were determined initially by manual inspection of sequence data, then confirmed by direct sequencing of PCR products from BAC DNA by USH2A cDNA primers. The initial determination of intron/exon boundaries was by manual inspection of sequence data. All boundaries were found to be consistent with the invariant intronic AG and GT features that flank the beginning and end of the majority of vertebrate exons, respectively. The contig assembly program AUTOASSEMBLER (ABI) was used on all sequence data generated. Seventeen separate sequence contigs from an input of 174 sequence files were generated and used to confirm the genomic organization. The total length of the 17 contigs was 33.6 kb.

The Usherin gene is encoded by 21 exons (fig. 1). Table 2 lists the sizes and starting positions of each of the 21 exons, the surrounding intronic sequence, intron sizes (where known), and GenBank accession numbers from the National Center for Biotechnology Information. Reliable genomic sequence extends 667 nt 5′ from the start of the cDNA sequence. The last genomic contig extends 605 bp 3′ from the published end of USH2A.

A significant sequence difference between the published cDNA and genomic DNA concerns the ATG translation start site. A stretch of five thymidines was previously observed to start at cDNA position +5 relative to the ATG start site; this was found to be six thymidines on all other genomic-based templates. The additional thymidine removes five amino acids from the beginning of the cDNA's longest open-reading frame (fig. 2A). A reexamination of the USH2A cDNA primary sequence data for this region revealed that the cDNA does contain six thymidines, in complete agreement with the genomic sequence. The poly-A tail of the cDNA sequence corresponds to a 20-nt imperfect poly-A stretch interrupted by 2 Gs in the genomic sequence (fig. 2B). It is likely that the true 3′ end occurs downstream of this imperfect poly-A feature. A potential polyadenylation signal (aataaa) was found 340 nucleotides 3′ from the end of the poly-A feature. The SignalP program (SignalP V1.1 World Wide Web Prediction Server, Center for Biological Sequence Analysis) predicted that the first 31 aa of Usherin would be a eukaryotic signal peptide (Nielsen et al. 1997). Therefore, the change in primary structure of Usherin from that previously reported (Eudy et al. 1998) does not alter the predicted role of this motif as a signal peptide for potential transport into the extracellular space. The sequence difference reduces the total length of the Usherin protein from 1551 to 1546 amino acids and results in a change of nomenclature for reported mutations; thus, the 2314delG (E772FS) mutation that we previously reported has become 2299delG (E767FS).

Mutations

Table 1 provides the primers, designed from intronic sequence, that were used in the mutation testing we describe here. Exons 1, 2, 5, 13, 17, and 21 were surveyed for mutations by direct sequencing. Exons 3, 14, 16, and 19 were screened by heteroduplex analysis, and heteroduplex-positive samples were sequenced to identify the mutation. To compare the sensitivity of the two methods for the detection of USH2A mutations, exons 4, 6–12, 15, 18, and 20 were screened by both methods.

Seven sequence changes—373A→G, 802G→A, 779T→G, 3883C→T, IVS15+35A→G, 4510–4511insA, and 4457A→G—were found by sequence analysis. Twenty-four sequence changes were observed first by heteroduplex analysis. In instances where PCR DNA was subjected to both heteroduplex analysis and sequencing, the detection rate for heteroduplex analysis was 92% of that of direct sequencing. Heteroduplex analysis did not detect any mutations not observed by direct sequencing. Thus, direct sequencing is more sensitive than the heteroduplex method at picking up mutations.

Table 3 lists 31 different nucleotide changes that were observed in this sample and gives their frequencies in the patients with Usher II and unaffected control patients; table 3 also identifies the restriction enzymes that we used to verify the base changes. The frequency of USH2A mutations detected is 58 of 114 USH2A alleles (51%). The number of USHIIa cases homozygous for mutations is 10, the number of compound heterozygotes is 10, and the number of cases with only one identified mutant allele is 18. USH2A mutations were not detected in 19 patients with Usher II. These proportions lead us to estimate that we have detected 69% of the true USH2A mutations and that 4 of the 19 individuals with no observed mutation harbor a change in USH2A that remains to be identified. This further leads us to conclude that ∼26% (15/57) of the families with USH2A in this sample are not linked to USH2A. This value corresponds well with an estimate, obtained by heterogeneity analysis of linkage marker data (Pieke-Dahl et al. 1997) 18% of cases are not linked to 1q41.

The locations of the 17 pathologic mutations on a motif-based map of the Usherin protein are summarized in figure 3. The protein has one laminin type VI domain, 10 laminin-like epidermal growth factor domains, and four fibronectin type III domains. It is of interest that the three probable pathologic missense mutations are limited to the domain VI region. All the other pathologic mutations we observed produced nonsense codons, frame shifts of mRNA translation, or splice mutations.

Nonsense Mutations

Seven stop mutations distributed throughout the Usherin gene were observed. None were observed in the series of 95 controls. All individuals had a typical type II phenotype.

The 779T→G (L260X) mutation was observed in a Swedish type II family as a compound heterozygote with the 2299delG mutation discussed below. A G→A transition was seen at bp 1227 in exon 7 (W409X) in two apparently unrelated Dutch families, one homozygous and the other a compound heterozygote with 1256G→T (C419F). In exon 10, a 1696C→T (Q566X) mutation was observed in one Swedish family as a heterozygote. A 1876C→T (R626X) mutation in exon 11 was observed in an American family of mixed northern European ancestry that also carried a 2299del mutation. The 2541C→A (C847X) mutation in exon 13 was observed with the 2299delG allele as a compound heterozygote in an American family with African heritage. There were two stop mutations in exon 18; one is an unusual double transversion 3840-41GA→CT (MR1280IX), and the other is 3883C→T (R1295X). The double transversion occurred in an American family with European ancestry. R1295X was seen in the heterozygous state in one sibship of a large Swedish family with three involved sibships; for both families, the companion mutation was not detected.

Frameshift Mutations

Five mutations cause frameshifts with premature termination. The first, a four-base insertion in exon 6, 921-922insCAGC (H308FS), was observed in two compound heterozygotes. The families were reportedly unrelated, but both were of English-Scottish ancestry. One family segregated the 2299delG mutation on the other allele; the other inherited a novel 2052A→G (Q684Q) maternal allele that could create a possible cryptic splice acceptor site 81 nt 3′ of the normal splice acceptor site inside exon 12. One additional 2052A→G allele from an unrelated USHIIa patient was observed; this mutation was not observed in any of the 95 control samples by heteroduplex analysis. A 1679delC (P560FS) mutation occurred in combination with 2299delG in an individual who has mixed northern European ancestry. One consanguineous individual from Venezuela was homozygous for 2898delG (T967FS), causing a frameshift starting at codon 967. A deletion of 2 bp, 4338-9delCT (Q1447FS), was observed in a French Acadian family with a well-documented Usher II phenotype. Although the parents of the affected children are believed to be distantly related, the children are predicted to be compound heterozygote, with the companion mutation still undetected.

The most common Usherin mutation was 2299delG (E767FS). We have observed this mutation in 31 of 192 alleles. The increase in the total number of patients tested for this mutation as well as portions of exons 12, 13, 14, and 20 is due to our previous work (Eudy et al. 1998). This mutation occurred among USHIIa patients with a variety of ancestries, but all were European. The single exception was the 2299delG/C840X compound heterozygote with African heritage, discussed above.

Splicing Mutations

To evaluate the 2052A→G (Q684Q) mutation, we used the FITCONSENSUS program from the GCG package to calculate quality scores for the authentic exon 12 acceptor site and the potential cryptic site created by the mutation with the eukaryotic 3′ splice acceptor consensus table (Mount 1982). The normal splice site quality score was 42.89; the Q684Q mutation score was 42.28. This mutation was absent from the 95 control samples; however, its potential to produce aberrant splice products has not been tested by reverse transcriptase–PCR methods. The other nonpathologic alterations from table 3, namely 2109T→C (D703D) and 4371G→A (S1457S), were also checked for potential splicing effects by FITCONSENSUS. No significant changes in the quality scores for the natural splice sites in these exons was observed. The 4371G→A does creates a new AG site in exon 20, but the quality score calculated from FITCONSENSUS at this potential cryptic acceptor site was 37.67, well below that of the natural site (44.28). In addition, restriction digests of PCR products from the 95 normal controls indicated that 2109T→C and 4371G→A are both present (table 3).

Missense Mutations

Seven different amino acid changes were observed in the sample groups. Three of these, 373A→G (T125A), 1931A→T (D644V), and 4457A→G (K1486R), are clearly polymorphic, as indicated on the basis of their frequency in the USHIIa population or the normal controls (table 3). In addition, both alleles of the 373A→G polymorphism were observed from BAC clone sequences of exon 2. The 802G→A (G268R) change was observed in a Swedish family in which inheritance of the gene was maternal. We were not able to test for the presence of G268R in the control samples, because this change did not produce a heteroduplex pattern and no restriction sites were altered by it. However, exon 5 DNA sequence from amplified gorilla DNA was homozygous for the rare human variant (unpublished observations), and so this base change is not considered likely to be pathologic.

The three missense mutations in USH2A, 956G→A (C319Y), 1042A→C (N346H), and 1256G→T (C419F) (fig. 3), all are located within domain VI and are absent from 95 control samples (fig. 3). The first, C319Y, was observed to be homozygous in the proband of a Hispanic American family with no family history of consanguinity. N346H was observed in the same large Swedish family described above that segregated 1295X. The N346H mutation was also observed in an American family as a compound heterozygote with 2299delG. The C419F mutation was observed exclusively in five Dutch families. Three of five Dutch families were simple heterozygotes, two were compound heterozygotes, and one carried 2299delG and the other W409X.

To evaluate the potential impact of these substitutions on the Usherin protein, a multiple sequence alignment was generated by 32 protein sequences with type VI domains, to identify conserved residues. A portion of the aligned set of laminin type VI globular domains from representative proteins and the locations of the three Usherin missense mutations is shown in figure 4. The C319Y and C419F mutations involve cysteine residues absolutely conserved in the type VI domains of these proteins. The N346H mutation occurs at a nonconserved residue, but the alignment identity of the asparagine residue with human laminin γ1, Caenorhabditis elegans laminin-like protein-1, human netrin-1, and C. elegans UNC6 suggests that this residue is also functionally conserved.

Disequilibrium

A 1931A→T polymorphism causes a nonconservative amino acid change, D644V, in the third laminin-like epidermal growth factor (L-EGF) domain of Usherin. The polymorphism was observed in the USH2a population as two D644V homozygotes and seven D644V heterozygotes, for a gene frequency of .096. In the control population, one homozygote and 16 heterozygotes were found, resulting in a gene frequency of .094. The C419F mutation appears to be in disequilibrium with D644V, because all five Dutch C419F alleles carried in cis the rarer D644V allele. None of the 95 control samples carried the C419F mutation, as determined by digestion of exon 7 PCR products with ApoI. Another sequence variant, IVS7+36-39delGATT, was observed to be in complete disequilibrium with all the USH2a patients who have the D644V allele. Unfortunately, we were unable to find USH2A mutations segregating with the 6 IVS7+36-39delGATT/D644V alleles in USH2a patients who did not have the C419F mutation. Concerning the IVS7+36-39delGATT allele among the control group with the D644V allele, BsrI digests of exon 7 PCR products showed that 4 of 16 D644V heterozygotes do not carry an IVS7+36-39delGATT allele, and the one normal control D644V homozygote was heterozygous for IVS7+36-39delGATT.